If your chatbot gives wrong prices, first prevent it from making unapproved quotes. Use one current price source for fixed offers, remove conflicting documents and route variable or uncertain requests to a person. A longer prompt alone is not a dependable pricing control.

Customers may treat a confident answer as a business commitment, even when the system generated it incorrectly.

Reproduce the error
Save the question and incorrect answer without exposing private customer information. Identify whether the bot used an old document, misunderstood the product, mixed currencies or invented a price when none was available.

Test similar questions and paraphrases. A one-off correct answer after editing the prompt does not prove the issue is fixed.

Establish one approved source
For fixed products, maintain a price table with product identifier, currency, applicable taxes or delivery exclusions, validity and approval owner. Archive outdated versions and remove duplicate conflicting entries from the bot's knowledge.

If prices depend on quantity, location or specifications, the assistant should collect those details and explain that a confirmed quote will follow. Do not let it estimate unless the business has explicitly approved the method and disclosure.

Separate retrieval from calculation
Where possible, use a controlled lookup or business rule for a price, then let the assistant explain the result. Validate that the lookup returned the correct product and required fields before sending the answer.

Any calculation should use approved inputs and be checked. A system that cannot determine a delivery charge should say so rather than silently treating delivery as free.

Design the uncertain-answer path
Example: 'I can share the published price for [fixed item]. Your request needs a custom quote because [reason]. I can collect the details and have our team confirm it.'

Give the team a task containing the request and what is missing. Avoid making the customer repeat the conversation. Define who handles pricing disputes or corrections if an earlier answer was wrong.

Test before reopening quoting
Create tests for unavailable items, old promotions, alternative currencies, vague product names, bundled services and attempts to push the bot into ignoring its instructions. Check that the system stays within approved pricing behavior.

After each price change, rerun the relevant tests. Restrict who can edit pricing knowledge and keep a change record. Monitor actual conversations for errors, with appropriate access controls.

A chatbot can still answer opening hours, service descriptions and booking questions while quoting is restricted. Keep the useful functions and repair the risky one.

CMH Consultancy can help audit knowledge sources and design a pricing handoff. Bring the approved price list and examples of incorrect answers so the repair is grounded in the actual failure.

Discuss your workflow with CMH Consultancy or call/WhatsApp +1 868 472 0763.

Related reading

How to Know If Your Business Is Ready for AI Automation.